RE: Phytotaxa iris paper

---

Hi all,
 
The at least "23 genera" paper is a phylogenetic study based on plastid DNA segments, which presents the evidence to support splitting Iris into all these genera.  It does not convince me in the slightest, though I find the study very well presented, fascinating, and educational.  I have not read the "rainbow" paper yet, but in it some of these genera (and a few more) are officially described.  It is interesting to note how closely this paper parallels the ideas that Rodionenko has been publishing since the 50's.  I should note that I think Rodionenko makes certain traits too important (such as the development of bulbs), and that results in too many genera.  However, I do find his classification to be mostly much more natural than the classification that the AIS and most Americans & western Europeans have been following for some time now.
 
My opinion - there is nothing in the "23 genera" paper that proves it necessary to divide Iris up into so many genera, and I find the whole concept premature and excessive.   However, the evidence can be used as support for this.  It can also be used to back up other scenarios, such as - at the other extreme - a system with only two large genera and a few stragglers that are of uncertain placement.  In fact, depending on interpretation of the data, it can be used to argue for 3, 5, 7, or how ever many genera you choose to read into the data.  Personally I would favor two genera, and perhaps two or three smaller genera that may not fit into either group.  Basically still the "Beardless" (Chamaeiris is the oldest generic name available - not Limniris) and the [mostly] "Bearded" Iris (genus Iris).  The bulk of the Crested Iris, Juno, Pardanthopsis, and Belamcanda would end up in the genus Iris, and the rest in genus Chamaeiris.
 
A good case can be made for a Crested Iris genus (Evansia), to include the Nepalensis group and the true Asian crested Irises (japonica, tectorum, etc.);  for a genus Juno; for a genus Iris (for all the Aril/Bearded types).  If those three are divided, then the genus Belamcanda would have to be recognized too, to include I. domestica and I. dichotoma (which are way too closely related to one-another for me to personally accept as being members of two distinct genera - but of course they still can be).  Iris verna comes out as a long branch sister to the ones listed above, and would have to be left out of all four genera if the data is interpreted strictly (I would personally put it in genus Iris (tentatively) under this scenario.
 
Following through with the same logic.  There is a major two-way split in the Beardless Iris, with one group (this would become Chamaeiris) including the bulbous species and the Spurias, and with the other group (this would become Limniris) including most of the rest.  I would not take it any further than this.  However, in this study, Siphonostyllus (with two species I. unguicularis and I. lazica), Lophiris (two American Crested species I. cristata and I. lacustris), and the other [mostly] American Crested Iris group (can be called Dielsiris or Rodionenkoa; including tenuis, missouriensis, longipetala, and gracilipes) all fall outside of the Beardless group and would have to have their own genera.
 
So, in this scenario there would be ten genera, well-supported, based on the same data.  I would find this arrangement much more natural and realistic than having 20-some genera of Irises.
 
However, it should be remembered that this is plastid DNA (assuming I read the article correctly), which is inherited nearly entirely maternally.  A study based on nuclear DNA would probably give somewhat different results, because that DNA has a different history, and is more likely to reflect past hybridization events (of which I would suspect there were very - very many).  This would probably especially affect the long-branch taxa (those rooted near the base of the "tree").  Many taxa might be rooted differently in such a study, showing somewhat different results.  In fact different runs of the same data in the same computer programs will show some variance in the results.
 
This last bit has only a little to do with the data presented here, it is interesting to note that many of the more "primitive" long-branch taxa share the trait of bearing a crest on the falls.  This is a trait that is scattered through almost all of the major groupings of Irises, and is likely a primitive ancestral condition in the Irises.  It's presence or absence, and differences in definition of what exactly qualifies as a crest, has created a lot of confused classifications in the past.  The Reticulatas, some of the Louisiana Iris, some of the more primitive Spurias, the series Chinensis, the Junos, etc. all show a crest in some stage of development, and I. brevicaulis (a Louisiana) even has a beard.  So, it should not be worried about too much that some of the traditional classifications (rather unnatural in many cases) are going out the window.  I never accepted most of the Crested Iris as being "beardless" anyway,  it was always obvious that most of them were more closely related to Bearded Iris and Junos than to most of the "Beardless" species.
 
If anyone is interested, I can send an outline drawn from Rodionenko's recent work (up to 2008) that sums up his classification (outdating the 1961 classification that is presented in the SIGNA publications).  I also have a modified version of the trees in the "23 genera" paper that shows some alternate classifications based on the same data.
 
Well, I guess that's enough of my unsolicited comments.
 
Best wishes to all,
 
Dave Ferguson
Albuquerque, New Mexico